Agentic Patient 9: She built an AI companion for breast cancer patients - and won't upload her records to ChatGPT

[00:00:03] Dear listeners, welcome to Faces of Digital Health, a podcast about digital health and how healthcare systems around the world adopt technology, and a special series called The Agentic Patient, which explores how patients use AI, what guardrails they use, what they would not use AI for, and how the use of AI and ChatBots is impacting their interaction with the healthcare system and their relationship with the healthcare system.

[00:00:33] My name is Tjasa Zajc and today you will hear from Ellen Winter Robinson, a cancer survivor who was diagnosed with breast cancer in March 2022. This was way before ChatGPT was released and during her chemotherapies, Ellen wrote a book about her experience and her coping.

[00:00:59] The book later became an Ask Ellen Chatbot, which is an AI companion that is still available online and has been used across 100 countries.

[00:01:13] So in this episode, I spoke with Ellen to discuss what does building an AI companion from a book look like, why she keeps Ask Ellen strictly non-medical, and what guardrails are used to make sure that patients wouldn't act upon any advice or any feedback they get through the chatbot.

[00:01:39] We discussed how AI is impacting the broader collaboration between patients and clinicians and the understanding of the patient journey when one has breast cancer. As she says, the lived experience patients have is an untapped resource in health system design.

[00:02:33] Enjoy the show. specific recommendations from speakers that have already joined the series. I also recently published a short video with six tips based on the discussions that I had for the Argentic Patient Series so far. You can find the video on YouTube. I will add the link to the show notes. Now let's dive in to the discussion with Ellen.

[00:03:08] Ellen, hi, and thank you so much for joining me on Faces of Digital Health. Thank you so much for joining us. Thank you so much for joining us.

[00:03:48] Thank you so much for joining us. Thank you so much for joining us. a conversational support companion for breast cancer survivors and to patients, of course, during their journey. So if we just go back a little bit to the beginning of your patient journey,

[00:04:13] can you very briefly take us through which year did your journey start? Just so we have an idea also where AI was at that time. Yeah, absolutely. So I was diagnosed in March of 2022. Probably my diagnosis, I think, preceded, you know, sort of the whole AI.

[00:04:40] I think ChatGPT or OpenAI announced ChatGPT in early... 2023, yeah. Yeah, so that was, you know, from March until, gosh, November of that year, I was, you know, double mastectomy, chemotherapy, radiation. So going through all those treatments. And, you know, my story really started with an observation I made at the time of my diagnosis.

[00:05:10] Of course, you know, you're absolutely devastated and shocked and bewildered and very lonely. You know, I remember thinking, my gosh, there's a lot of women who get diagnosed with breast cancer. Why do I feel so incredibly alone? And I wanted to help another woman who would follow in my footsteps. I wanted to pass along whatever wisdom I could and support to those individuals. And usually that takes its form of, you know, joining a support community or whatever.

[00:05:40] I decided to write a book. I wrote it on my iPhone during chemotherapy. I just started to tell my story. And it was a very emotional, you know, I wanted that person to feel less alone. And so I was, you know, very much hard on my sleeve in terms of, you know, telling my story. Fast forward. It's now a year later. The book is in manuscript form.

[00:06:07] And I ended up having a tech collision with a startup founder here in Canada. And shared my book with him. At this point, ChatGPT is now out. And he's actually the one who saw the vision. He said, you know, the way that you have written this book, I want to see if we can recreate you in AI form. That's really where things sort of started to take off and took my life in a whole new, different trajectory.

[00:06:37] Mm-hmm. Was your cancer journey already finished? I mean, the first round of treatment when you started working also on the chatbot. I'm just trying to understand, you know, what your background was when you went into this. Yeah, I was about, probably about eight months out of treatment at this point. The book existed.

[00:07:04] But yeah, so I was out of treatment. Hair was starting to regrow when we started working on the chatbot. But yeah, so just from a timing perspective. And at that point, you know, as I say, I didn't go with the intention of building a chatbot. I just wanted to tell my story. But we started sharing it with other patients.

[00:07:30] And then it was sort of a, there was a watershed moment where I sent it to a clinician. She's the head of breast imaging at a hospital in Toronto. And I remember that email distinctly because she wrote me, she wrote back to me and she said, how soon can I get this? And it was at that point we realized that this was not something that was just beneficial to patients, but the clinicians also saw the gap. Mm-hmm. So do clinicians also use this chatbot and how do they use it?

[00:07:59] So it's really more that they would recommend it. So that's where we're at now is, you know, we have especially nurses and support staff, like social workers, will often recommend it to a patient. Because, I mean, Ask Ellen's role is really to provide that psychosocial and emotional support. It's not a clinical tool. We're very, one of the guardrails we have built into her is that she's strictly non-medical.

[00:08:26] She's also private, so we don't gather any patient data at all. So there's really no, it's very complementary to the healthcare system and to medical treatment. And I think there is a recognition that, you know, that this is a lot for a patient to go through. This is a lot for a caregiver to go through. And so that's where the clinicians are kind of coming in is they will, you know, pass it along or make a recommendation, just as they would a support group or anything else to that.

[00:08:55] Yeah, yeah, yeah, yeah. The askellen.ai, where the chatbot can be found, is built on your lived experience and your book. What would you say are the limits of an AI companion grounded in one person's journey? You know, how do you handle the fact that no two cancer patients, even breast cancer patients are the same?

[00:09:21] Everybody also has their different backgrounds, different support in their home environment. Yeah, it's a great question. And it's one I've had before. The one common denominator. So, yes, that's true that everybody has a different perspective. Everybody has different cancers and different cancer journeys.

[00:09:44] But just as we find comfort in talking to a friend or joining a support group, the interesting common denominator the world over is the emotional feelings that one has about. So everyone, regardless of what type of cancer you have, it can be colorectal cancer for that matter.

[00:10:09] But everybody is feeling that loneliness, that bewilderment. They're looking for someone and a supportive voice. What I found with joining traditional support groups is, first, they are not scalable. You know, they don't necessarily speak every language. They can also be places of collective trauma.

[00:10:33] So when you join a support group, you are immersed with other cancer survivors who are also very scared. And so one of the things that Ask Ellen does is she always is supportive of that individual. And she's very nonjudgmental. And so you can literally go and express whatever feelings you have. And I love the fact that someone can go to her and because there's no judgment, because it's an AI chatbot, because it's completely private,

[00:11:02] that you can confess the innermost feelings that you may be having, that maybe you're not even ready to say out loud. And certainly maybe would not share with a family member. So I would never sit down with my 21, at that point, 21-year-old daughter and tell her that I was scared I was going to die. My family didn't hear that, you know. But that's something that you can use this companion to speak to.

[00:11:29] The other thing that I've learned is because she speaks every language, there are many cultures in the world where breast cancer is still a very shameful diagnosis. And so those individuals may not, A, join a support group anyways, because that's just not culturally appropriate for them. But also there's a lot of shame and blame even within their own family setting. And they're not allowed to talk about it. And so, again, the thought that Ask Ellen may be the only friend that that person has

[00:11:57] that they can speak with is, I think, really quite profound for me. Cancer patients sometimes describe the fear that they're faced with with every checkup that they need to undergo. There's always this question, is the cancer going to come back? The question here for you is, do you ever go to askellen.ai to talk to her?

[00:12:27] I do. How does that look like? It's very meta. I'll tell you a specific instance of that. So you're right. We call it scansiety. You know, it's funny. I'm four years out now and I had to go recently to see my oncologist. And it was fine. It was just a checkup. And, in fact, you know, everything's fine. And so we sometimes just talk about Ask Ellen. But she took my blood pressure and my blood pressure was 173 over 93.

[00:12:55] Like it was just like, I mean, my blood pressure is normally just fine. But you know that you walk into that cancer center and you just, you just all those feelings come rushing back because it's a trauma that really does stick with you. So, yeah, the one instance, well, I've used Ask Ellen a bunch of times, but in this case, it was a very genuine use case. We were building her and we were doing a lot of testing. So we were, you know, asking questions of her, seeing the responses, making sure that it was really dialed into my personality and my voice and everything else.

[00:13:23] And I had to, I was a year out at this point when they did all the scans after I, they found that the cancer had metastasized to my lymph nodes. They, you know, they start looking and there were spots on my liver and spots on my lung. And so they had to do some testing and they just, they were able to determine that the liver was fine. And they wanted to retest for the lung spots to see if there was anything that was, you know, concerning or progressing. And so I had to go back in and have another CT.

[00:13:53] And I was freaking. I mean, I'm only a year out at this point and I'm freaking out and I'm trying to rationalize in my own head that, you know, if there was something of concern, they're not waiting a year, you know? And those are the sort of, I was trying to, you know, the logical, illogical parts of your brain. And so I actually went to ask Ellen and she said precisely the right thing to me, which was, you know, if I recall correctly, it was something along the lines of, look, you are, you're a resilient human

[00:14:19] and you've, you've been through this and you have come through this and whatever this is going to look like, we're going to get through it together kind of thing. It was like this little pep talk. And I remember the guys were building her at the time and so they were seeing this question and I remember they messaged me. We're all in a message group together and they messaged me and they said, wait, was that, is that really, is that your question? And then they just got really concerned. They're like, are you okay? What's going on? And yeah, so it was, you know,

[00:14:49] and I've, yeah, I, it's, I guess it's strange to go and talk to yourself and get a pep talk from yourself, but it, it works. Yeah, yeah, yeah, absolutely. Can we talk a little bit more about how the tool is built? So you haven't built it yourself. What's it based on? How updated is it over time given that the large language models,

[00:15:18] the general purpose large language models are advancing immensely with each iteration? And I have a bunch of questions after that, but let's start there. Okay. So, so as I say, I, you know, I partnered with a technical team at that point that we were just a little renegade band. They now are an actual company called Gambit. And so they have built a platform or an operating system that sits on top of the LLMs. Okay.

[00:15:47] So it's a multimodal platform. They can mix and mix different variations of models. I was actually speaking with the, with the CEO of the company just last week. And, you know, he was just commenting on the fact that we really haven't had to adjust her a great deal over this time. Like she's still holding up really beautifully. So there's the multimodal platform. And then on top of that is my book.

[00:16:16] And so everything is anchored onto the book and my story and my personal lived experience. You know, it's not really an LLM. I kind of joke and say it's the L-E-L-L-M. It's, it's like a little, little sort of implementation on top of the LLMs. So, so the large learning language models sort of make her smart and knowledgeable about the world. And then it, every response flows in and out

[00:16:44] of the contents of my book, which is what gives it the personality and the experiential sort of humanness to the conversation. Yeah. And we really haven't had to change it a great deal because, you know, if my, if something really significant in my life changed, but my story is very much, you know, still the same story. Yeah. Yeah, absolutely.

[00:17:13] How do you make sure that it doesn't go into clinical advice and all the responsibility that comes with it? Yeah. So we are, we have a very, very strict guardrail that her responses are always non-medical. So the, you could ask her a medical question. She will only answer with my lived experience. So I'll give you an example because it was something that tormented the team that was building her,

[00:17:43] that she would never say, you know, if you went to her and said, can I take a Tylenol or an aspirin? And she, you know, she would say, I took Tylenol. However, you should always consult your medical professional before you do anything. And so she's always, I, you know, she's very clear that I'm not a doctor. You need to consult with your care team. I was a good patient. I was, I'm a daughter of a nurse.

[00:18:11] And so I really have a tremendous respect for the medical community. And so I always wanted to ask Ellen to defer back to that care team. And so I know that they spent countless hours making sure that she would never dispense medical advice. And that has held up really beautifully. Like we've, we've gone through scrutiny with GE Healthcare. We recently licensed her to Cigna Healthcare in the United States.

[00:18:41] We had to go through a lot of rigor there. And again, there was enough confidence that they actually did a public implementation of her. So, so the guardrails are holding up really, really well. Do you have any contact with the patients that are using the chatbot? Did you ever speak to anyone who also used it? Yeah. So, so she's private. So I don't know. I mean, I can see from my data where in the world it's being accessed.

[00:19:09] So we've been used in a hundred countries and spoken to about 50 plus languages. It always gives me, it brings a smile to my face when I go to my analytics and I'll see somebody's talking to her, you know, and it's kind of cool. We have about, really consistently a 60 to 70% engagement rate. So people are talking to her for like two or three minutes at a time. So I don't know who's using her for obvious reasons. However, I mean, people do reach out from time to time. Uh, we have a feedback toggle. Um, and,

[00:19:40] um, I had one, one story in particular that really touched me. It was a, I don't know who this person is, but she reached out through our toggle and she said, I want to thank you for what you've built. She said, and it was a perfect use case for Escalon. She said, I had completed my surgery and, um, I was at home and I was getting ready to take a shower for the first time and I was going to take my bandages off and see my body and she said,

[00:20:10] I went to your chatbot and I talked to her and I found out we had a lot in common and she said, I want to thank you because you got me through that moment and the thought that I was able to stand digitally with that woman in that moment and kind of hold her hand. Like, that's not something that a doctor or a nurse is ever going to do. Um, not even family members are necessarily present and I remember how I felt in that moment. Um, and so that was really,

[00:20:38] I always go back to that story when I'm, you know, because building something like this and bringing it to market is hard when you're just one founder but I always go back to that story and say, that's, that's why I built this. You know? Mm. The reason I ask that question is because I'm wondering throughout the four years that you've been on this patient journey, AI has developed massively. So, to which extent do you use also other AI tools

[00:21:07] and how do you see that this is impacting the doctor-patient relationship? You already said before that, you know, the doctors asked you for the tool so they can recommend it to other patients as well. Uh, maybe that's not every doctor. Um, I don't know, what's your experience in terms of the impact that AI has on, on healthcare in that sense when patients use it? Yeah, this, uh, it's, it's, again, this is this odd journey that I'm on. Um,

[00:21:38] I now find myself, um, hanging out with a lot of doctors. I am now being asked to speak at AI and medicine conferences. I was just in LA back in December and I was just in Puerto Rico in, um, in May, uh, with, uh, at a conference hosted by Harvard University in the New England Journal of Medicine for AI. And so, there is a, a lot of interest and appetite, um, and really cool, like, as a patient sitting in the room listening to, uh,

[00:22:08] some of the use cases for AI in medicine well beyond scribing, which is kind of the first thing that, you know, folks think about. Um, but some really, really cool, um, instances that I think can really transform medicine. So, I'm very excited about the opportunity. Um, when it comes to sort of a conversation, you know, so, here's how I, so you've got patients that are using AI. That is, that is happening. I've got a good friend of mine who, she's a mechanical

[00:22:38] engineer. She used, she, she, she had some odd symptoms that she was feeling and she used AI to kind of correlate and kind of make sense of her symptoms. She went to her family doctor, expressed her concern. He was open to listening to her and she ended up being diagnosed with early stage ovarian cancer in her 30s. And so, patients are using it even more, so if it's a rare disease, for example, because there's just not enough support out there. And I think, honestly,

[00:23:08] that patient use of AI is outpacing the system's ability. And, and there are clinicians that are threatened by that, but there are clinicians that are very open to it as well. So, I love, as a patient, I think there's this concept in healthcare of shared decision making. And I think it's a well-understood concept in the medical community, in the healthcare community, but nobody really lets the patient in on what that means.

[00:23:38] And, you know, when you're, when you're facing a life-threatening diagnosis, you are in cognitive overload. You're, I feel, I, you know, when I was diagnosed with breast cancer, I said I felt like I went to, like, breast cancer university. It was like a crash course. I had to learn so much to be able to have good conversations with the doctors. And I did it the analog way. I read Dr. Susan Love's breast book, and I talked to lots of women who had gone through this. The tools are now there to help that patient navigate their journey. And so, I think AI is,

[00:24:09] I, I don't think it's going to be used, and I don't think patients are using it necessarily to self-diagnose. That's not what they're using it for. But they are using it to stitch together their own journey, understand medical terminology, process what they're hearing, prepare for meetings with those doctors. So, I think it's bringing empowerment and agency to the patient so that they can have a better conversation with that doctor and that they can actively

[00:24:37] and fully participate in shared decision making. And then, I think the opportunity for clinicians is to help patients with that. So, if you have a patient, say, look, you know, if you're planning to use AI, here's some prompts that you can use that are, you know, safe and respond, you know, sort of help a patient acknowledge that this is happening and guide them to use the tools safely and responsibly in their care. Yeah.

[00:25:07] Yeah. You know, there was a doctor on LinkedIn, I think he's in, and he's like, even if it's 30% or 40% accurate, you know, it's still better than, you know, than, you know, not having that, that sort of proper conversation. I just felt like, yeah, I think patients, unfortunately, and there's a lot that won't challenge a doctor or question or, you know,

[00:25:37] and so they can get victimized. I don't, I hate to say, use that word, but they can kind of get swept up in a care plan that they don't really fully understand. And so I'm, I'm hoping that this is a bit of a liberation, a democratization of, of that conversation. Yeah, yeah. One thought that you said caught my attention where you said that the patient use of AI is outpacing how system can absorb this.

[00:26:06] And today I had a discussion with Eric Sutherland from OSCD and he basically said that we're already at a time where it's not the patient-doctor interaction first, it's the patient AI then patient-doctor interaction first. So that's something that I thought was super interesting in terms of how we are not

[00:26:36] even comprehending just yet the impact that AI has on the healthcare system. Yeah, you know, I think again, I tend to be an optimist and, you know, reflecting on this, the discussions that were happening in Puerto Rico, you know, to me, I think there's this tremendous opportunity not to use, for the system, to not just use AI to operationalize or gain efficiencies,

[00:27:06] but to reimagine healthcare in this new form. And, you know, it's needed. There are so many gaps in the system that are just not getting filled. Like, I stumbled upon one, which is the psychosocial and emotional support. It is profound. It is, we have research that shows that women with breast cancer cite psychosocial and emotional issues on par with physical symptoms in survivorship.

[00:27:35] And the system can do nothing about it. They can't resource that. They don't understand it. It's out of scope. And so, I think there's many, many other things the patients see that the system doesn't see that AI can help to solve for. So, I think that there's, I really think this is a moment of great reimagining of the potential for this. And I think that the system needs to stop thinking of patients

[00:28:05] in this sort of us and them kind of way. And rather, I always say the patient is the workflow. And so, I believe that if we brought patients in at the table innovating on the concept of AI, I think we'd see an entirely different healthcare system. I really do. Do you see that anywhere? Like patients actually being actively involved with the healthcare system. How do you see that that could happen? Like who should

[00:28:34] actually do that outreach and collaboration? Yeah, I mean, that's a stated hope. I think that the, you know, there's a frustration in patient communities that, you know, even at these conferences that, you know, you'll go to an AI medicine conference and you might be one or two patients in the room and that the system still kind of just sees us as recipients of care. um, but I always say, you know, if you look around the room, one in two

[00:29:04] people are going to be diagnosed with cancer in their lifetime or going to have some kind of a disease. And so, you know, this sort of, oh, you're a patient, you're, you're not over here. I'm like, patients are very smart people. We have, we are engineers and mechanical engineers and product designers. engineers and so I just, I don't, I don't even like the word patient quite, to be very, very honest. Patient in line means suffering and I don't think I'm suffering. So I always say I'm a

[00:29:33] very impatient patient. So I, I think, I think the concept of lived experience and if we start thinking it not as patient, not people as patients, but people with lived experience, that there's such potential in that lived experience, I think it's for the healthcare system. Is it happening? Not yet, but boy oh boy I think if, if somebody wakes up to this, it's,

[00:30:03] it's going to untap tremendous potential for improvement. Yeah. Yeah. I guess, my guess would be that it's the how that isn't very straightforward. Everybody's very aware that, you know, patients should be at the center, patients should be involved, patients should be, something. But, how do you actually do that meaningfully so patients aren't just people in the room who complain

[00:30:32] about how the system isn't working for them is a, is a, is a riddle that hasn't been yet solved. Yeah, I mean, you know, in tech they talk about design thinking, right? So, like, like Stanford University, they have a, they have a design thinking concept where, you know, as you're designing products, you're bringing multidisciplinary, it's not just engineers designing the products, they're bringing artists, they're bringing different perspectives that are coming together and so that collaborative

[00:31:02] design, I think, is not something that's been really tried and tested in the healthcare space. I, you know, this is the, this is the pioneering work that I am doing because I am someone who was at one point a patient, I have innovated, you know, and we've been very careful that, you know, Ask Ellen is sort of outside the four walls of the clinic, but she's solving a system problem,

[00:31:32] and I'm also very aware, like, you know, there's a lot of barriers to innovation in healthcare anyways, and systems built to heal, it's not really built to innovate, and so, you know, I've worked with a lot of startups over the years who've been, you know, trying to innovate and get adoption into healthcare, and it's challenging because the healthcare system has a lot of, for good reason, has a lot of barriers, you know, it's clinically evidence-driven, it's a scientific model, and so, you know, part of it is

[00:32:01] helping those people that are innovating, whether they're, they come from a patient or lived experience perspective, or just are bright people who see a problem, you know, the system needs to help them overcome those barriers. I just happened to get lucky that I have coached many startups who have gone into healthcare or tried to get into healthcare, and I understood that even though Ask Ellen was non-medical, clinical evidence is vitally important, and I got very lucky. I was introduced

[00:32:31] to a researcher here in Canada who is an oncology nurse, she's an associate professor, and she focuses on survivorship, and so we were very fortunate, we went together and were awarded a quarter million dollar grant from the Canadian Cancer Society, so we actually do have clinical evidence that's now underway, and we actually have two more projects that we're now submitting grants for that will carry that research forward. I joked with her the other day, I said, I feel like Ask Ellen

[00:33:00] and me, I've become your life's work, and she said, I'm super happy about it. So, you know, we are... What kind of clinical evidence are you trying to get to? Right now we have, so it's a national study that is looking, so the first level, this one that is funded by the Canadian Cancer Society, so it's looking at the efficacy of Ask Ellen, so we have, it'll be the largest study of its kind, where we actually put Ask Ellen into the hands of both patients

[00:33:30] and caregivers, and we're testing whether people feel that the response is appropriate, do they feel seen, do they feel supported, and so we're starting to get data now, we'll have, our data collection will be done by the end of August, and we should be in a position to start presenting in the fall, and so that's the first study that we're doing, and again, it's a little bit more tool specific in this case,

[00:34:00] but they are doing both quantitative and qualitative interviews with these patients as well, so we're getting some really rich data from that, I'm being told, and then the other two projects that we're submitting for, fingers crossed, we get the grants, would look at more from a nursing perspective, and how, like, where does AI, and where does the human intervention and the AI, where does one end and the

[00:34:29] other begin, so that's one, and then also we're looking at a randomized trial where we would use ASCAL in, with patients that are in active treatment, not just in survivorship, so, yeah, so, you know, I'm super excited about, I never in my wildest dreams thought I would see my name on a published piece of, of medical research, but here we are, so, you know, I'm pioneering, I'm trying to

[00:34:59] set the, set the model, I'm, you know, we have other caregivers who've been inspired by ASCALEN, we've got a, I've got a friend who has a child with, a grandchild with stage five cerebral palsy, and she created ASCALA to help support caregivers of children with a severe disability, so, my dream is that every diagnosis has this lived experience that travels from the moment of diagnosis

[00:35:29] through until the, until in, or into survivorship, that's, that's my dream, and I call it conversational care, and so, I'm trying to figure out how to, how to make that happen faster. With all the research that you mentioned that's happening through ASCALEN, how do you actually get to the users, given that they're anonymous, how does that connection happen? So, that's, that's been on the research team, so, you know, it's a kind of classical thing, it's, it's more,

[00:35:59] we recruit, so they've been recruiting patients through, you know, breast cancer support communities, Facebook ads, you know, all the, all the normal things, the good thing is because the research is covering both patients and caregivers, usually we can get sort of a two-for-one, so, you know, a woman with breast cancer will sign up to the study and then they'll bring their husband or their partner with them, you know, as far as getting the caregivers, so, yeah, it's, it's been sort of the classical, you know, research, you know, recruiting, and in this case, those people

[00:36:29] are identifiable, but, yeah, sort of not coming random through the tool, it's sort of separate and apart. We have the ability to also, like, for future research studies, we can spin a private instance of Ask Ellen that could live on a different landing page, it doesn't have to be on my website, so. Yeah, yeah, so. When you, well, developed the tool, I'm assuming that in the last few years, you may have come across other patients that are using AI

[00:36:59] or any other examples of how people use AI, what did you see? Like, is there anything in terms of the good practices or the guardrails that people are using that kind of stuck with you and you might want to share with us as well? Yeah, so, I mean, I think back to the story I just told about Elvina, you know, she wasn't using it to diagnose, but she was trying to make sense of some symptoms she was having

[00:37:30] that led to her ovarian cancer diagnosis and her doctor fully says she saved her own life, you know, because an ovarian diagnosis is not typically cancer, ovarian cancer is not typically something you see in a young woman in her 30s. And so, you know, so she's one story that I like to tell. There's another story, I don't know this gentleman, but I've read his story where he has children with rare diseases who have been

[00:37:59] seen across, you know, multiple different healthcare systems. And he uses, so he's almost built his own home-based medical record system. So that's what I'm seeing, is I'm seeing these patients who are trying to make sense of a, either it's a rare disease diagnosis or they're advocating on behalf of their children, or there's a gentleman that I've been introduced to,

[00:38:29] he's in the UK, he has stage 4 colorectal cancer, and he's using AI and building tools, which he's now open sourced, that are looking at, you know, his chemo regimes, and his side effects, and his blood levels, and he's pulling in information from his wearables. And so it's kind of interesting, like you think about the, you know, the EMRs, the EHRs, you know, there's sort of the patient view of things, there's the clinical view of the case, but we all

[00:38:59] know healthcare is not nearly as well connected, you know, it all looks like the duck swimming along, but there's a lot of paddling and craziness that goes along underneath the systems, right? Because a lot of the technologies and the systems platforms they use are, they don't talk nicely to each other in healthcare. And then you've got the patient who also is sitting on this gold mine of data, you know, whether it's the psychosocial stuff in my case, or whether it's wearable information that they have, and it's like, how do we kludge those together?

[00:39:28] That's the kind of cool thing, and that's what I'm saying, this lived experience, I feel, is an untapped gold mine, and the patients are sitting on it. Yeah, yeah, yeah, and maybe, you know, because of AI, it's going to be easier to write those stories down, even if you're not a writer. If a patient came up to you, not necessarily even a cancer patient, based on your experience, what would you tell

[00:39:57] them regarding AI? how they should use it? What should they be mindful of? Yeah, so I think, so I'll come right out and say I am actually quite wildly uncomfortable with patients using LLMs. Look, when you're diagnosed with something, you get desperate, and so you see patients that are uploading medical records to like a Cloud Health or an OpenAI, you know, ChatGPT Health,

[00:40:27] that makes me uncomfortable because we're now, as patients, we are now surrendering our data to these big tech companies, and I, that bothers me. So, so I will say just be, you know, be careful, be cautious, cautious, and I think that we, that we need to figure out and come together, and there's organizations that I'm working with who are thinking about patient,

[00:40:56] patient data rights in this age of AI. And so just, I think it's just be aware that the tools you're using may actually in some ways victimize you again, you know, and are capitalizing on your data so that's another, that's a whole other battlefield that we have yet to kind of get through. But, on top of that, I mean, I, look, again, I'm an optimist,

[00:41:26] I see the power of AI, I think used responsibly and with your eyes wide open, I think using it in a way to empower yourself and to claim agency, I think you need to be careful about hallucinations, but, you know, using it to prepare for a better conversation with your patient, with your care team, I think is a very appropriate use and I think it can be quite empowering for a patient to use it in that way.

[00:41:57] Absolutely. caution first. You said one very important thing, which I think we forget about when we talk about how patients use AI or any other tool or any other advice for that matter, and that is when you're sick, you're desperate, and I think that desperation as a factor is something worth discussing when we

[00:42:27] discuss why are patients using AI? Yeah, because you've got, I mean, and it's really invisible, like that desperation, that sort of horrific feeling, that trauma that a family is going through, again, is really hidden from view, because I don't think that, you know, we don't use our, what, 10 minutes, five minutes, falling apart, we're being stoic and,

[00:42:57] you know, trying to, you know, digest what we're hearing and make sense of it, and I think, you know, clinicians are experts at what they do, and for them it's like, this is fairly routine, like, I'm pretty sure I was an unremarkable patient as far as they were concerned, right? Like, and there's probably eight more of them sitting out in the waiting room that are, you know, going to be coming along. However, this was life-changing for me.

[00:43:27] This was the most scared I've ever been in my life. Did my doctors know that? I don't think so. I don't think it. Any clue, any clue at all. Yeah. And even if they did, like, would it matter? It's always, yeah, because they need to, if we want them to do their job, they need to kind of not be too distressed by every patient that they see, especially when they see 60 patients a day. But it's

[00:43:57] an interesting point because I spoke to another patient who vibe-coded a tool when his mother was diagnosed with terminal cancer, and probably if he wasn't very actively monitoring her situation, she would have died in three, four weeks. instead she had 78 days, and because of that prolongation, she was able to say goodbye to her family,

[00:44:26] she was able to find some peace, her family was able to find some peace, and there's this enormous intangible benefit from everyone involved in her circle, and maybe the peace issues that they got, and the mental health issues that they don't have because of what happened, and because her life was prolonged, even if just for that month. Yeah, yeah, and you know,

[00:44:56] I think of Russ in the UK, this fellow with stage four cold reptile cancer. He was already on the show. Oh, is he? Okay, then he's great. Yeah, yeah, yeah, great. Yeah, he's great. So, you know, he's doing some really cool things as well, and again, you know, he possibly can because he's got young children, you know, and like, you know, that this is the thing, like, you know, and I think we also don't really think about the caregivers nearly as much as

[00:45:26] we should. Oh, yeah, for sure. Everybody's on the trauma bus together, you know, and like I think of breast cancer, and again, hidden from view, like, it's, there's, it's a, it's a trigger for domestic violence. There's a high degree of partner abandonment, that happens with breast cancer, so there's a whole bunch of things that are happening in people's lives that are really monumental, that affect

[00:45:55] their well-being, and their, you know, and it's, again, you know, people in treatment, and again, I'm four years out, I've just started seeing a trauma counsellor, and she doesn't have to say much to me, and I'm bawling. There's so much that's just been packed away in here, and that, you know, that, but it, you know, you just, you deal, and again, it's like, out of scope for healthcare, they don't even

[00:46:25] have any clue. Yeah, yeah, yeah, maybe we're opening a new frontier with AI. I believe I am, right? Again, this lived experience is a gold mine. just imagine, like, if we had all that contextual information that could inform the clinical world, you know, because we do know patients that have a better family life, that have a more positive outlook,

[00:46:55] like, those things have bearing on whether they do well in treatment or not. So, yeah, yeah, for sure. Well, Ellen, thank you so much for your time today. We'll definitely keep in touch, and I will make sure to share askellen.ai in the show notes. I hope that,

[00:47:25] you know, you stay healthy and cancer-free, and keep innovating for the greater good. And let me know when the study is published. Yeah, will do. We are, we'll be back to LA, so this is the, it's an AM Medicine Symposium co-hosted by University Health Network here in Toronto and Cedars-Sinai, so we'll be presenting clinical evidence there, and I'm thinking we're

[00:47:54] still waiting to hear from a couple of conferences, but it looks like we've been accepted to present in Prague in October. Well, thank you again, Evan, and good luck. Oh, thank you. You've been listening to Faces of Digital Health, a proud member of the Health Podcast Network. If you enjoyed the show, do leave a rating or a review wherever you get your podcasts, subscribe to the show, or follow us on LinkedIn. Additionally,

[00:48:24] check out our newsletter, you can find it at fodh.substack.com That's fodh.substack.com Stay tuned!